DNA Methylation Epitypes Sort 1,262 De Novo AML Patients and Catch Mutation Mimics That Genetic Testing Misses

From 649 array profiles to a 1,262-patient classifier

The Ohio State group defined the epitype framework five years ago. In a 2021 Genome Research paper, Brian Giacopelli, Oakes and coauthors ran genome-wide Illumina methylation arrays on 649 AML patients and found 13 clusters through unbiased analysis, most of them tied to a recurrent mutation or mutation combination, with a minority independent of any single genetic driver. That paper also reported that epitypes with stem-cell-like methylation had inferior overall survival, that a STAT-motif signature travelled with FLT3-ITD, and that methylation signatures were stable at relapse in most patients. Giacopelli is second author on the new Blood paper.

The 2026 study applies the framework to 1,262 patients with de novo disease, roughly double the discovery set, and asks a clinical question the earlier paper did not: does the epitype add anything once a hematologist already has the karyotype, the mutation panel and the patient's age and blood counts? The abstract does not name the trial cohorts that supplied the samples. The author list includes the Clara D. Bloomfield Center for Leukemia Outcomes Research at Ohio State and leukemia investigators at Wake Forest, Northwell, Emory, the University of Maryland, the University of Alabama at Birmingham and Dana-Farber.

Patients who look like FLT3-ITD without FLT3-ITD

Most epitypes were dominated by one genetic alteration. The finding the authors lead with concerns the exceptions. Within an epitype, some patients lacked the cardinal alteration yet shared its methylation pattern, and the paper labels them "genetic alteration-like." For CEBPA bZIP, FLT3-ITD, core-binding factor and KMT2A-rearranged AML, and for what the abstract calls "other abnormalities," those alteration-like patients "displayed outcomes similar to patients with actual cardinal alterations." In practical terms, a patient with no detectable KMT2A rearrangement whose methylome says KMT2A behaved like a KMT2A patient.

The FLT3-ITD analysis goes a step further. The group derived a signature of hypomethylation at STAT binding sites that was enriched in FLT3-ITD cases and called it the STAT hypomethylation signature. SHS positivity identified patients with inferior outcomes and, per the abstract, added to the prognostic weight of FLT3-ITD itself. Since FLT3-ITD already moves patients within the European LeukemiaNet 2022 genetic risk groups and determines midostaurin or quizartinib eligibility, a methylation marker that separates FLT3-ITD patients into better and worse subsets would have an immediate use, if it validates.

A model layered on genetic, demographic and clinical markers

The prognostic claim rests on machine learning modeling in which the methylation signatures were added to genetic, demographic and clinical markers. According to the abstract, the combined model "significantly added" to prediction of complete remission, relapse and overall survival. The abstract does not state the model type, the size of the improvement, the validation strategy or whether the comparison was made against the ELN 2022 categories as such or against the component variables. Only the abstract was available for this report; the full paper is behind the American Society of Hematology's subscription wall, and the numbers behind the word "significantly" could not be checked.

The authors' summary is that methylation signatures "capture patients who mimic cardinal genetic mutations providing additional prognostic information that may be used to in concert with standard genetic markers."

Why This Matters to the APO|APE Reader

The 2021 Genome Research paper used full Illumina arrays, and a 1,262-patient classifier that can be run on clinical samples implies a targeted assay rather than a genome-wide one, so the assay format, turnaround and specimen requirements in the full paper will decide whether an epitype call can sit alongside the cytogenetics and NGS panel at diagnosis or remains a research readout. The nearest precedent is the Ohio State group's own methylation-iPLEX test for CLL developmental subtypes, published in Blood in 2019, which never became a routine send-out. Two of the four mimic categories, KMT2A-rearranged and FLT3-ITD, now carry approved targeted drugs, revumenib and the FLT3 inhibitors, and the paper stops short of claiming that a methylation mimic should be treated as a carrier, which is the question a molecular tumor board will ask first.